Analysis of weekend effect in severe acute liver injury: A nationwide database study

• Main Authors: Albert Do, Ysabel C. Ilagan‐Ying, Tamar H. Taddei
• Format: Article
• Language: English
• Published: Wiley 2020-03-01
• Series: Health Science Reports
• Subjects: acute liver failure; health services; inpatient; mortality; nationwide/national inpatient sample; predictors
• Online Access: https://doi.org/10.1002/hsr2.139

Abstract Background and Aims Severe acute liver injury (ALI) can lead to poor outcomes without timely management. Comparatively worse outcomes in various severe, emergent conditions have been attributed to reduced hospital resources experienced by patient weekend admissions, a phenomenon termed “weekend effect.” To date, a weekend effect has not been studied in severe ALI, an emergency also necessitating timely management. We aimed to evaluate such an effect in this condition by analyzing a large national inpatient database in the United States. Methods We analyzed the Nationwide/National Inpatient Sample (NIS) 2000 to 2014, the largest inpatient, all‐payer database in the United States (US), containing sociodemographic, clinical, patient‐, and hospital‐level data. We identified severe ALI using International Classification of Disease, 9th Revision diagnosis codes for acute/subacute hepatic necrosis (570) with encephalopathy (572.2). Our primary outcome was in‐hospital mortality. Using a full‐model approach for covariate selection, we performed multiple logistic regression modeling to assess for weekend effect and identify predictors of in‐hospital mortality. Results We identified 15 762 eligible hospitalizations, with 12 182 (77.3%) having complete covariate data. This sample comprised 53.3% males, 69.3% White race, and had an average (± SD) age of 55.0 ± 14.1 years. We utilized a full‐model approach for covariate inclusion but did not include patient transfer data due to limited availability. We observed no significant mortality differences in weekend admissions (OR = 1.06, 95% CI: 0.97‐1.15, P = 0.02). However, significantly higher mortality was associated with male sex, older age, Black or Hispanic race, Northeast US hospitalization, urban teaching status, and larger hospital size. Sensitivity analyses using multiple imputation datasets and transfer covariates did not change our results. Conclusion We did not observe a weekend effect of in‐hospital mortality for weekend admissions for severe ALI, but our overall diagnosis ascertainment yield was low—indicating that lack of accurate documentation for the etiology of severe ALI may be masking an effect. Additionally, our findings suggest that racial differences and hospital‐level characteristics in the context of severe ALI may be associated with varying outcomes, regardless of admission day, which warrants further research.